HOLE: a program for the analysis of the pore dimensions of ion channel structural models

A method (HOLE) that allows the analysis of the dimensions of the pore running through a structural model of an ion channel is presented. The algorithm uses a Monte Carlo simulated annealing procedure to find the best route for a sphere with variable radius to squeeze through the channel. Results can be displayed in a graphical fashion or visualized with most common molecular graphical packages. Advances include a method to analyze the anisotropy within a pore. The method can also be used to predict the conductance of channels using a simple empirically corrected ohmic model. As an example the program is applied to the cholera toxin B-subunit pentamer. The compatibility of the crystal structure and conductance data is established.

Molecular dynamics simulations of isolated transmembrane helices of potassium channels

In the middle of the S6 helix in voltage-gated potassium channels there is a highly conserved Pro-Val-Pro motif, while the equivalent M2 helix of inward rectifier potassium channels contains a conserved glycine residue in a comparable position. The structural implications of these conserved motifs are of interest given the evidence that S6 and M2 are components of the lining of their respective pores. Multiple sequence alignment and TM helix prediction methods were used to define consensus regions for S6 and M2. Ensembles of 50 structures for each helix were generated by simulated annealing and restrained molecular dynamics. Time-dependent fluctuations of S6 and M2 were investigated by long time scale molecular dynamics simulations on representative members of each ensemble carried out in vacuo in the presence and absence of a hydrophobic potential that mimics a lipid bilayer. The results are discussed in terms of the structural basis of the kink in S6 and M2 and of a putative functional role for flexible helices as "molecular swivels."

The pore domain of the nicotinic acetylcholine receptor: molecular modeling, pore dimensions, and electrostatics

The pore domain of the nicotinic acetylcholine receptor has been modeled as a bundle of five kinked M2 helices. Models were generated via molecular dynamics simulations incorporating restraints derived from 9-A resolution cryoelectron microscopy data (Unwin, 1993; 1995), and from mutagenesis data that identify channel-lining side chains. Thus, these models conform to current experimental data but will require revision as higher resolution data become available. Models of the open and closed states of a homopentameric alpha 7 pore are compared. The minimum radius of the closed-state model is less than 2 A; the minimum radius of the open-state models is approximately 6 A. It is suggested that the presence of "bound" water molecules within the pore may reduce the effective minimum radii below these values by up to approximately 3 A. Poisson-Boltzmann calculations are used to obtain a first approximation to the potential energy of a monovalent cation as it moves along the pore axis. The differences in electrostatic potential energy profiles between the open-state models of alpha 7 and of a mutant of alpha 7 are consistent with the experimentally observed change in ion selectivity from cationic to anionic. Models of the open state of the heteropentameric Torpedo nicotinic acetylcholine receptor pore domain are also described. Relatively small differences in pore radius and electrostatic potential energy profiles are seen when the Torpedo and alpha 7 models are compared.

Simulation of voltage-dependent interactions of alpha-helical peptides with lipid bilayers

Pore formation in lipid bilayers by channel-forming peptides and toxins is thought to follow voltage-dependent insertion of amphipathic alpha-helices into lipid bilayers. We have developed an approximate potential for use within the CHARMm molecular mechanics program which enables one to simulate voltage-dependent interaction of such helices with a lipid bilayer. Two classes of helical peptides which interact with lipid bilayers have been studied: (a) delta-toxin, a 26 residue channel-forming peptide from Staphylococcus aureus; and (b) synthetic peptides corresponding to the alpha 5 and alpha 7 helices of the pore-forming domain of Bacillus thuringiensis CryIIIA delta-endotoxin. Analysis of delta-toxin molecular dynamics (MD) simulations suggested that the presence of a transbilayer voltage stabilized the inserted location of delta-toxin helices, but did not cause insertion per se. A series of simulations for the alpha 5 and alpha 7 peptides revealed dynamic switching of the alpha 5 helix between a membrane-associated and a membrane-inserted state in response to a transbilayer voltage. In contrast the alpha 7 helix did not exhibit such switching but instead retained a membrane associated state. These results are in agreement with recent experimental studies of the interactions of synthetic alpha 5 and alpha 7 peptides with lipid bilayers.

Structure and orientation of the mammalian antibacterial peptide cecropin P1 within phospholipid membranes

Cecropins are positively charged antibacterial peptides that act by permeating the membrane of susceptible bacteria. To gain insight into the mechanism of membrane permeation, the secondary structure and the orientation within phospholipid membranes of the mammalian cecropin P1 (CecP) was studied using attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectroscopy and molecular dynamics simulations. The shape and frequency of the amide I and II absorption peaks of CecP within acidic PE/PG multibilayers (phosphatidylethanolamine/phosphatidylglycerol) in a 7:3 (w/w) ratio (a phospholipid composition similar to that of many bacterial membranes), indicated that the peptide is predominantly alpha-helical. Polarized ATR-FTIR spectroscopy was used to determine the orientation of the peptide relative to the bilayer normal of phospholipid multibilayers. The ATR dichroic ratio of the amide I band of CecP peptide reconstituted into oriented PE/PG phospholipid membranes indicated that the peptide is preferentially oriented nearly parallel to the surface of the lipid membranes. A similar secondary structure and orientation were found when zwitterionic phosphatidylcholine phospholipids were used. The incorporation of CecP did not significantly change the order parameters of the acyl chains of the multibilayer, further suggesting that CecP does not penetrate the hydrocarbon core of the membranes. Molecular dynamics simulations were used to gain insight into possible effects of transmembrane potential on the orientation of CecP relative to the membrane. The simulations appear to confirm that CecP adopts an orientation parallel to the membrane surface and does not insert into the bilayer in response to a cis positive transmembrane voltage difference. Taken together, the results further support a "carpet-like" mechanism, rather than the formation of transmembrane pores, as the mode of action of CecP. According to this model, formation of a layer of peptide monomers on the membrane surface destablizes the phospholipid packing of the membrane leading to its eventual disintegration.

Engineering stabilized ion channels: covalent dimers of alamethicin

The peptide alamethicin forms channels with a variety of conductance states. Selective stabilization of a particular state should simplify the task of understanding conductance in terms of channel structure. We synthesized two different covalent dimers of alamethicin in which peptides were linked at their C-terminal ends by flexible tethers. Both dimeric peptides formed channels with conductances that matched those of alamethicin channels. Particular conductance states were selectively stabilized, however, with lifetimes up to 170-fold longer than the same states observed with monomers. In addition, tethering appeared to limit the size of the structures formed so that, even at higher peptide concentrations, a single predominant conductance state was obtained. We suggest this state corresponds to a channel made from six alamethicin molecules (three dimers).

Molecular dynamics simulations of water within models of ion channels

The transbilayer pores formed by ion channel proteins contain extended columns of water molecules. The dynamic properties of such waters have been suggested to differ from those of water in its bulk state. Molecular dynamics simulations of ion channel models solvated within and at the mouths of their pores are used to investigate the dynamics and structure of intra-pore water. Three classes of channel model are investigated: a) parallel bundles of hydrophobic (Ala20) alpha-helices; b) eight-stranded hydrophobic (Ala10) antiparallel beta-barrels; and c) parallel bundles of amphipathic alpha-helices (namely, delta-toxin, alamethicin, and nicotinic acetylcholine receptor M2 helix). The self-diffusion coefficients of water molecules within the pores are reduced significantly relative to bulk water in all of the models. Water rotational reorientation rates are also reduced within the pores, particularly in those pores formed by alpha-helix bundles. In the narrowest pore (that of the Ala20 pentameric helix bundle) self-diffusion coefficients and reorientation rates of intra-pore waters are reduced by approximately an order of magnitude relative to bulk solvent. In Ala20 helix bundles the water dipoles orient antiparallel to the helix dipoles. Such dipole/dipole interaction between water and pore may explain how water-filled ion channels may be formed by hydrophobic helices. In the bundles of amphipathic helices the orientation of water dipoles is modulated by the presence of charged side chains. No preferential orientation of water dipoles relative to the pore axis is observed in the hydrophobic beta-barrel models.

Water in channel-like cavities: structure and dynamics

Ion channels contain narrow columns of water molecules. It is of interest to compare the structure and dynamics of such intrapore water with those of the bulk solvent. Molecular dynamics simulations of modified TIP3P water molecules confined within channel-like cavities have been performed and the orientation and dynamics of the water molecules analyzed. Channels were modeled as cylindrical cavities with lengths ranging from 15 to 60 A and radii from 3 to 12 A. At the end of the molecular dynamics simulations water molecules were observed to be ordered into approximately concentric cylindrical shells. The waters of the outermost shell were oriented such that their dipoles were on average perpendicular to the normal of the wall of the cavity. Water dynamics were analyzed in terms of self-diffusion coefficients and rotational reorientation rates. For cavities of radii 3 and 6 A, water mobility was reduced relative to that of simulated bulk water. For 9- and 12-A radii confined water molecules exhibited mobilities comparable with that of the bulk solvent. If water molecules were confined within an hourglass-shaped cavity (with a central radius of 3 A increasing to 12 A at either end) a gradient of water mobility was observed along the cavity axis. Thus, water within simple models of transbilayer channels exhibits perturbations of structure and dynamics relative to bulk water. In particular the reduction of rotational reorientation rate is expected to alter the local dielectric constant within a transbilayer pore.

Molecular modelling of Staphylococcal delta-toxin ion channels by restrained molecular dynamics

Delta-Toxin is a 26-residue channel-forming peptide from Staphylococcus aureus which forms an amphipathic alpha-helix in a membrane environment. Channel formation in planar bilayers suggests that an average of six delta-toxin helices self-assemble to form transbilayer pores. Molecular models for channels formed by delta-toxin and by a synthetic analogue have been generated using a simulated annealing protocol applied via restrained molecular dynamics. These models are analysed in terms of the predicted geometric and energetic properties of the transbilayer pores. Pore radius calculations of the models demonstrate that rings of channel-lining residues contribute a series of constrictions along the pore. Electrostatic properties of the pores are determined both by pore-lining charged side chains and by the aligned helix dipoles of the parallel helix bundle. Molecular dynamics simulations (100 ps) of delta-toxin models containing intra-pore water were performed. Analysis of the resultant dynamics trajectories further supports the proposal that alternative conformations of pore-constricting side chains may be responsible for the observed conductance heterogeneity of delta-toxin ion channels.

Water-mediated conformational transitions in nicotinic receptor M2 helix bundles: a molecular dynamics study

The ion channel of the nicotinic acetylcholine receptor is a water-filled pore formed by five M2 helix segments, one from each subunit. Molecular dynamics simulations on bundles of five M2 alpha 7 helices surrounding a central column of water and with caps of water molecules at either end of the pore have been used to explore the effects of intrapore water on helix packing. Interactions of water molecules with the N-terminal polar sidechains lead to a conformational transition from right- to left-handed supercoils during these stimulations. These studies reveal that the pore formed by the bundle of M2 helices is flexible. A structural role is proposed for water molecules in determining the geometry of bundles of isolated pore-forming helices.

Modelling packing interactions in parallel helix bundles: pentameric bundles of nicotinic receptor M2 helices

The transbilayer pore of the nicotinic acetylcholine receptor (nAChR) is formed by a pentameric bundle of M2 helices. Models of pentameric bundles of M2 helices have been generated using simulated annealing via restrained molecular dynamics. The influence of: (a) the initial C alpha template; and (b) screening of sidechain electrostatic interactions on the geometry of the resultant M2 helix bundles is explored. Parallel M2 helices, in the absence of sidechain electrostatic interactions, pack in accordance with simple ridges-in-grooves considerations. This results in a helix crossing angle of ca. +12 degrees, corresponding to a left-handed coiled coil structure for the bundle as a whole. Tilting of M2 helices away from the central pore axis at their C-termini and/or inclusion of sidechain electrostatic interactions may perturb such ridges-in-grooves packing. In the most extreme cases right-handed coiled coils are formed. An interplay between inter-helix H-bonding and helix bundle geometry is revealed. The effects of changes in electrostatic screening on the dimensions of the pore mouth are described and the significance of these changes in the context of models for the nAChR pore domain is discussed.

Transbilayer pores formed by beta-barrels: molecular modeling of pore structures and properties

Transmembrane beta-barrels, first observed in bacterial porins, are possible models for a number of membrane channels. Restrained molecular dynamics simulations based on idealized C alpha beta templates have been used to generate models of such beta-barrels. Model beta-barrels have been analyzed in terms of their conformational, energetic, and pore properties. Model beta-barrels formed by N = 4, 8, 12 and 16 anti-parallel Ala10 strands have been developed. For each N, beta-barrels with shear numbers S = N to 2N have been modeled. In all beta-barrel models the constituent beta-strands adopt a pronounced right-handed twist. Interstrand interactions are of approximately equal stability for all models with N > or = 8, whereas such interactions are weaker for the N = 4 beta-barrels. In N = 4 beta-barrels the pore is too narrow (minimum radius approximately 0.6 A) to allow ion permeation. For N > or = 8, the pore radius depends on both N and S; for a given value of N an increase in S from N to 2N is predicted to result in an approximately threefold increase in pore conductance. Calculated maximal conductances for the beta-barrel models are compared with experimental values for porins and for K+ channels.

Modelling membrane proteins using structural restraints

Here we present a procedure for modelling membrane proteins which employs molecular dynamics simulations incorporating target restraints derived from low-resolution structures alongside distance restraints derived from mutagenesis data. The application of the modelling procedure to the closed conformation of the pore domain of the nicotinic acetylcholine receptor is described. This domain is formed by a parallel bundle of five M2 helices. Each M2 helix is kinked due to cumulative distortions of backbone (phi, psi) values. The central region of M2 may adopt a more distorted conformation. This would enable a ring of conserved leucine residues (one from each M2 helix) to pack together, occluding the central pore and thus preventing ion permeation. Molecular dynamics simulations on isolated helices that kink formation is not an inherent property of M2.

Structural features of isolated M2 helices of nicotinic receptors. Simulated annealing via molecular dynamics studies

The nicotinic acetylcholine receptor is an integral membrane protein and a ligand-gated cation channel. It has stoichiometry alpha 2 beta gamma delta, the subunits arranged symmetrically around an approximate five-fold axis. Five M2 helices, one from each subunit, form a parallel helix bundle surrounding a central pore. Simulated annealing via restrained molecular dynamics (SA/MD) has been employed to generate ensembles of isolated M2 transmembrane helices. Four ensembles of two different M2 helix sequences, M2 delta and M2 gamma, have been generated by SA/MD. The ensembles differed in their treatment of electrostatic interactions. Analysis of the simulated structures showed that intra-helical H-bonds were more strongly conserved in the C-terminal (and more hydrophobic) segment of M2 helices. Conformations of polar sidechains have been analyzed, placing particular emphasis on EK (and QK) pairs at the N-termini of M2 delta (and M2 gamma) helices. Conformations of EK sidechain pairs were obtained for the high resolution structures in the protein database in order to guide our analysis of simulated structures. Serine and threonine sidechain conformations in the M2 models also have been determined. Implications of studies of isolated M2 helices for models of the intact pore region of the nicotinic receptor are discussed.

Ion channel formation by synthetic analogues of staphylococcal delta-toxin

Ion channel formation by three analogues of staphylococcal delta-toxin, an amphipathic and alpha-helical channel-forming peptide, has been evaluated by measurement of ionic currents across planar lipid bilayers. Replacement of beta-branched, hydrophobic residues by leucine and movement of a tryptophan residue from the hydrophilic to the hydrophobic face of the helix does not significantly alter ion channel activity. Removal of the N-terminal blocking group combined with the substitution of glycine-10 by leucine changes the single channel properties of delta-toxin, without altering macroscopic conductance/voltage behaviour. Truncation of the N-terminus by three residues results in complete loss of channel-forming activity. These changes in channel-forming properties upon altering the peptide sequence do not mirror changes in haemolytic activity. The results lend support to the proposal that channel formation and haemolysis are distinct events. Channel properties are discussed in the context of a model in which the pore is formed by a bundle of approximately parallel transbilayer helices.

Packing interactions of Aib-containing helices: molecular modeling of parallel dimers of simple hydrophobic helices and of alamethicin

alpha-Aminoisobutyric acid (Aib) is a helicogenic alpha, alpha-dimethyl amino acid found in channel-forming peptaibols such as alamethicin. Possible effects of Aib on helix-helix packing are analyzed. Simulated annealing via restrained molecular dynamics is used to generate ensembles of approximately parallel helix dimers. Analysis of variations in geometrical and energetic parameters within ensembles defines how tightly a pair of helices interact. Simple hydrophobic helix dimers are compared: Ala20, Leu20, Aib20, and P20, the latter a simple channel-forming peptide [G. Menestrina, K.P. Voges, G. Jung, and G. Boheim (1986) Journal of Membrane Biology, Vol. 93, pp. 111-132]. Ala20 and Leu20 dimers exhibit well-defined ridges-in-grooves packing with helix crossing angles (omega) of the order of +20 degrees. Aib20 alpha-helix dimers are much more loosely packed, as evidenced by a wide range of omega values and small helix-helix interaction energies. However, when in a 3(10) conformation Aib20 helices pack in three well-defined parallel modes, with omega ca. -15 degrees, +5 degrees, and 10 degrees. Comparison of helix-helix interaction energies suggests that dimerization may favor the 3(10) conformation. P20, with 8 Aib residues, also shows looser packing of alpha-helices. The results of these studies of hydrophobic helix dimers are analyzed in the context of the ridges-in-grooves packing model. Simulations are extended to dimers of alamethicin, and of an alamethicin derivative in which all Aib residues are replaced by Leu. This substitution has little effect on helix-helix packing. Rather, such interactions appear to be sensitive to interactions between polar side chains. Overall, the results suggest that Aib may modulate the packing of simple hydrophobic helices, in favor of looser interactions. For more complex amphipathic helices, interactions between polar side chains may be more critical.

Ion-channel gating. Twist to open

Three-dimensional images of the open state of the nicotinic acetylcholine receptor have been obtained at 9 A resolution. Comparison with the closed state reveals the structural basis of channel gating.

Seven-helix bundles: molecular modeling via restrained molecular dynamics

Simulated annealing via restrained molecular dynamics (SA/MD) has been used to model compact bundles of seven approximately (anti)parallel alpha-helices. Seven such helix bundles occur, e.g., in bacteriorhodopsin, in rhodopsin, and in the channel-forming N-terminal domain of Bacillus thuringiensis delta-endotoxin. Two classes of model are considered: (a) those consisting of seven Ala20 peptide chains; and (b) those containing a single polypeptide chain, made up of seven Ala20 helices linked by GlyN interhelix loops (where N = 5 or 10). Three different starting C alpha templates for SA/MD are used, in which the seven helices are arranged (a) on a left-handed circular template, (b) on a bacteriorhodopsin-like template, or (c) on a zig-zag template. The ensembles of models generated by SA/MD are analyzed in terms of their geometry and energetics, and the most stable structures from each ensemble are examined in greater detail. Structures resembling bacteriorhodopsin and structures resembling delta-endotoxin are both represented among the most stable structures. delta-Endotoxin-like structures arise from both circular and bacteriorhodopsin-like C alpha templates. A third helix-packing mode occurs several times among the stable structures, regardless of the C alpha template and of the presence or absence of interhelix loops. It is characterized by a "4 + 1" core, in which four helices form a distorted left-handed supercoil around a central, buried helix. The remaining two helices pack onto the outside of the core. This packing mode is comparable with that proposed for rhodopsin on the basis of two-dimensional electron crystallographic and sequence analysis studies.